EIC Climate Change Technology Conference 2015

Transcription

1 Increasng the Lmts of Allowed Dstrbuted Generatons by Volt-VAR Control H. Ahmad 1 and J. R. Martí 1 1 Unversty of Brtsh Columba, Vancouver, BC, Canada Abstract Renewable energy resources are beng ntroduced to the grd, both at transmsson and dstrbuton systems level. A partcular challenge n ntegraton of renewable energes at dstrbuton systems s the ssue of voltage rse on the feeders. Due to ths techncal ssue, the penetraton level of dstrbuted generaton has been lmted n some areas to mantan the voltages wthn the acceptable ranges. In ths paper, a volt-var control strategy s proposed to ncrease the capablty of feeders to accommodate more renewable generatons wthout jeopardzng system operaton lmts. Ths strategy s targeted for near-real-tme operaton and s expected to mprove the overall system effcency. A 69-node test system s used to demonstrate the effectveness of the proposed algorthm. Keywords: dstrbuted generaton, voltage rse, voltage-var control Résumé Les ressources énergétques renouvelables sont ntégrées à la fos aux réseaux électrques de transport et de dstrbuton. Un déf partculer lors de l'ntégraton d énerges renouvelables à des réseaux de dstrbuton est la possblté de surtensons aux départs de lgne. En rason de ce problème technque, le nveau de pénétraton de la producton décentralsée a été lmté dans certanes zones pour mantenr les tensons à l'ntéreur des lmtes acceptables. Dans cet artcle, une stratége de contrôle volts-var est proposée afn d'augmenter la capacté d ntégraton de sources d énerge renouvelables des artères de dstrbuton sans compromettre les lmtes de fonctonnement du système. Cette stratége est destnée à fonctonner en temps quas réel et devrat amélorer l'effcacté globale du système. Un système de test de 69 noeuds est utlsé afn de démontrer l'effcacté de l'algorthme proposé. Mots clés : producton décentralsée, surtenson, tenson de commande-var 1. Introducton Due to the unsustanable and pollutng nature of fossl fuels, renewable energy resources (RER) have been sought greatly n the past decades. The efforts n fndng new sources of energy have led to numerous alternatve generaton methods, such as wnd, solar, tdal, geothermal, bomass, fuel cell, etc. Among those, wnd and solar power generatons have taken the lead thank to the emergng technologes n energy conversons. As was mentoned n a report from the US Department of Energy n 2010, twenty four states are expected to supply 10%-40% of ther load usng renewable energes by [1]. Despte the great benefts of replacng fuel-based generaton unts by renewable alternatves, the RERs cause many techncal ssues n power systems operaton and control [2]. In the transmsson systems, the problems stem from the three major characterstcs of RERs, as follows: 1) non-dspatchable; 2) ntermttent and uncertan; 3) ncompatble wth the conventonal system. Whenever power s avalable from a RER, t should be all absorbed by the grd. In other words, there s no control on the amount of avalable power from a RER. Ths makes t very 1

2 challengng for the system operator to coordnate other generaton unts n order to supply the varable load and mantan the load-generaton balance at all tmes. Ther ntermttent and probablstc nature of the RERs make t complcated to plan ahead of tme and mantan a hgh relablty for the system operaton. A large amount of system reserve s requred to account for the unpredctable changes n the power generaton from RERs [3]. The energy converson technology that many of the RERs adopt s substantally dfferent from the conventonal synchronous machnes used n thermal, hydro, CHP (combned heat and power), and nuclear power plants. For example, synchronous machnes have a large mechancal nerta and t makes the system reslent to many dsturbances. Many RERs, on the other hand, are connected to the network through a power electronc converter whch does not provde any mechancal nerta. In that sense, hgher penetraton level of RERs s translated nto a more vulnerable system wth smaller nerta [4]. In dstrbuton systems, smaller scale of RERs, called dstrbuted generaton (DG), are present. The type of problems these DGs may cause are, n part, dfferent from those at the transmsson systems. Dstrbuton systems were bult to operate n radal topology wth the power flowng from the substaton to the loads. All the control and protecton schemes were desgned based on ths assumpton. When DGs wth consderable sze are connected to a feeder, they create a reverse power flow back to the substaton and ncrease the short-crcut levels at dfferent parts of a feeder [5]. Besdes nterferng wth the protecton relays, hgh penetraton of DGs can cause a voltage rse problem. Each utlty has standards for the mnmum and maxmum range of voltages at the customer sdes to ensure hgh qualty power delvery to the customers. Accordng to the Canadan Standard Assocaton, the standard voltage ranges for dstrbuton systems (below 50 kv) are defned for normal and extreme condtons n [6]. These values are reproduced here n Table 1 for reference. Due to ths voltage rse problem, the penetraton level of DGs at some feeders s lmted [7], [8]. In ths paper, a control strategy s proposed to mantan the voltages wthn the standard range and mnmze power losses. Other methods have been developed n the lterature and are brefly revewed n the followng. The mpacts of hgh penetraton of photovoltac (PV) DGs on dstrbuton feeders were analyzed n [9]. The voltage rse and reverse power flows were shown n [9] to be among the mportant lmtng factors for hgh penetraton of PVs. A power curtalment algorthm was proposed n [10] to reduce the voltage rse effect durng the peak power producton of PVs. A coordnated power curtalment strategy was also proposed n [11] whch consders a droop for each PV unt to control the amount of curtalment per each unt. However, by curtalng the power generaton, the chance of producng clean energy s lost. The authors n [12] proposed addng a battery storage system to each PV unt to make the njected power to the network manageable. However, addng batteres causes addtonal costs and envronmental concerns. To avod batteres, a smple compensator was added to the PV converters n [13] to exchange arbtrary amount of reactve power wth the network. Ths method s a good soluton. However, t requres coordnaton wth other voltage- VAR control equpment, such as lne voltage regulators, capactor banks, dstrbuton statc compensators (D-STATCOM), etc. The ssue s partally addressed n [14] n whch a coordnaton strategy s proposed to control voltage regulators and battery storage systems when loads and PV generaton vares substantally durng the day. In [15], DGs are taken advantage of n provdng reactve power support to the grd. However, at the peak generaton nterval, the full capacty of Table 1 Standard voltage ranges for dstrbuton systems accordng to CSA Vmn (p.u.) Vmax (p.u.) Normal operatng range Extreme operatng range

3 DGs s usually used for actve power producton and there s no room for reactve power exchange wth the grd. For nstance, the maxmum output power of PVs usually occurs when the load s not at peak and, therefore, PVs cannot be reled on to provde VAR compensaton at ths tme, when t s most needed. The applcaton of D-STATCOM for voltage regulaton n dstrbuton systems wth hgh penetraton of PVs was studed n [16]. It was shown that by adoptng the D-STATCOM there s no need for actve power curtalment. However, the coordnaton of D-STATCOM wth other compensaton equpment was not consdered n [16]. In ths paper, an optmzaton problem s formulated for centralzed coordnaton of D-STATCOM and lne voltage regulator n a dstrbuton feeder wth hgh penetraton of RERs. Wthout loss of generalty, the RERs are consdered to be PVs. The proposed optmzaton routne reduces the network losses whle ensurng the voltage lmts and feeder s ampacty. Ths method s based on the lnear power flow (LPF) formulaton developed by the authors n [17]. Usng a lnear formulaton nstead of the conventonal nonlnear and non-convex equatons leads to a better performance n the soluton of an optmzaton problem. The rest of the paper s organzed as follows. In Secton 2, the proposed optmzaton strategy s descrbed. Smulaton results and test cases are gven n Secton 3. The man fndngs of ths study are summarzed n Secton Centralzed Coordnaton Algorthm The objectve of voltage-var control (VVC) n ths paper s to allow for hgher penetraton level of DGs. The constrants are the nodal voltage lmts, feeder s ampacty, lmts of compensaton devces, and lmts on tap postons of under-load tap-changng transformers (ULTC). In order to perform VVC n near-real-tme, load data and lne mpedances are needed. In a centralzed control mechansm, t s assumed that the communcaton nfrastructure s avalable between all the control equpment and the control centre. Also, the amount of actve power generaton from each sgnfcant DG s communcated to the control centre. In the followng, a bref descrpton of network modellng s provded. 2.1 Lnear power flow Loads are essentally voltage-dependent and the constant-power load model may not be vald for dstrbuton systems analyss. In ths paper, loads voltage dependence s modeled usng the followng equatons: P 2 = CZV + CIV (1) 2 = C ' Z I Q V + C V (2) n whch C, C, C, C are parameters determnng the voltage dependence of each partcular load; Z I Z I P and Q are the actve and reactve power of load n per-unt, respectvely; V s the termnal voltage n per-unt. DGs are modelled as constant-power loads wth negatve values. For constantpower load model, the followng parameters are used: CZ =- 1, CI = 2. Usng the proposed load model, and assumng small voltage angles n dstrbuton systems, a lnear power flow (LPF) formulaton was derved n [17]. The LPF equatons are as follows: 3

4 n re m å ( Gmk, Vk - Bmk, Vk ) = Ipm, (3) k = 1 n m re å ( Gmk, Vk + Bmk, Vk ) = Iqm, (4) k = 1 where G and B are the real and magnary parts of the modfed admttance matrx; n s the number re m of nodes; V and V are the real and magnary parts of the nodal voltages, respectvely. In the modfed admttance matrx, the mpedance parts of the loads are added to ts correspondng dagonal elements; the current njecton parts of loads appear n the rght-hand sde of (3) and (4) as I p and I q. 2.2 Under-load tap-changng transformer ULTCs are equpped wth tap changers. Tap changer can dynamcally adjust the transformer rato, normally wth dscrete steps, and regulates the voltage on ts secondary sde wth respect to a reference set pont. Tap poston, denoted by a, can be used to model the ULTC usng varable passve elements, as n [18]. Fgure 1 shows the equvalent crcut for the voltage regulator. Increasng the tap above 1 creates an nductve shunt at the prmary and a capactve shunt at the secondary. Ths ncreases the voltage on the secondary sde, whle potentally reducng the voltage on the prmary sde. The tap poston a s essentally an nteger quantty. However, t can be modelled as a contnuous varable and then rounded to the closest nteger value. By dong ths, t s not possble to guarantee the global optmalty of the soluton. Another approach s to model ths as an nteger varable and adopt mxed-nteger programmng technques to solve the optmzaton problem. It s well known that mxed-nteger programmng problems are computatonally ntense problems to solve and are not sutable for onlne applcatons. Therefore, a contnuous relaxaton of the problem s consdered here, whch provdes a fast, whle suboptmal, soluton usng nonlnear programmng (NLP) algorthms. The tap poston s bounded to ts maxmum and mnmum values: mn max a a a (5) where mn a and max a are the mnmum and maxmum tap postons. 2.3 Dstrbuton statc compensator t Fgure 1. Equvalent crcut of ULTC ( y : transformer short-crcut mpedance; a : tap poston). 4

5 A D-STATCOM s a member of the flexble alternatve current transmsson systems (FACTS). It conssts of a power electronc nterface, nductors (couplng transformer) and capactors. By controllng the nverter, t s possble to exchange reactve power wth the grd. It has a sgnfcant advantage over statc VAR compensator (SVC). The termnal voltage does not affect the amount of VAR that can be njected to the network by D-STATCOM, whle t changes quadratcally wth the termnal voltage n SVCs. In ths paper, t s assumed that the amount of reactve power exchange can be remotely controlled n D-STATCOM [19]. The nternal actve power losses n D- STATCOM s assumed to be neglgble. The amount of VAR exchange s constraned usng the followng equaton: mn max s Qs Qs Q (6) mn max n whch Qs and Q s are the mnmum and maxmum lmts on the VAR exchange of D- STATCOM, respectvely. 2.3 Volt-VAR control algorthm The coordnated voltage and VAR control has four major parts, as shown n Fg. 2. All the data s collected at the control centre and a decson s made based on the receved measurements. In a more realstc case, a state estmaton algorthm s used to estmate the data for the unmontored nodes and remove outler measurements. Dstrbuted generatons are assumed to be nondspatchable, meanng that the full amount of avalable power wll be njected to the network. A closed-loop near-real-tme algorthm for VVC s proposed here, as shown n Fg. 3. In ths algorthm, an optmzaton routne s called anytme a consderable change happens n the network. A consderable change s defned as large devaton n network voltages from ther last recorded values when the prevous settngs were sent out. It can be quantfed by settng the followng threshold lmt: V - V V (7) now old th where V now s the most recent voltage measurement; V old s the voltage measurement record from the last tme any settngs were sent out; V th s a threshold set for trggerng the algorthm to modfy the controllers. The value of ths threshold should be chosen n a way to always mantan the voltages wthn acceptable ranges whle not trggerng the algorthm too often and mechancally stress the equpment by frequently alterng ther settngs. The optmzaton routne can be set to one of the followng cases Mnmzng losses The total power losses n the network can be defned as: re re 2 m m 2 mk[ m k m k ] Ploss = å G, ( V - V ) + ( V -V ) (8) m< k 5

6 Control Centre Dstrbuted Generatons Voltage Regulators D-STATCOM Fgure 2. Communcaton between control centre and the control equpment. Collect Measurements V now -V last >V th No Yes Run Optmzaton Routne Infeasble Soluton? Yes Mnmze Infeasbltes No Dspatch Unts Fgure 3. Closed-loop Volt-VAR control algorthm for near-real-tme. The optmzaton problem conssts of mnmzng (8) subject to the power flow equatons of (3) and (4), tap poston lmts n (5), lmts of D-STATCOM output n (6), and the followng constrants: mn max V V V (9) I I (10) max where V s the voltage magntude at node and I s the current magntude n branch Flattenng voltage profle 6

7 In order to flatten the voltage profle, t s desrable to have all the voltages close to a certan value, e.g., 1 per-unt. In order to express ths statement n a mathematcal formula, the followng voltage ndex s defned: n 2 Vj Vtar (11) j 1 n whch V tar s the target value for the voltage profle. Usually, t s desrable to have all the voltages close to 1 per-unt. By mnmzng subject to the same constrants as n Subsecton 2.3.1, voltage profle of the network can be brought close to the targeted value Mnmzng nfeasbltes In some of the real stuatons, t s not always possble to satsfy all the constrants of the optmzaton problem. In such cases, the problem s nfeasble. The best practce n those cases s to mnmze the nfeasbltes. Let us assume that some of the nequalty constrants are nfeasble: f ( x) 0, IF (12) where IF s the set of nfeasble constrants. In such cases, the objectve of the VVC would be to mnmze the volatons. New varables, z, are ntroduced to the problem and the constrants n (12) change to the followng constrants: f ( x) z, IF (13) The objectve functon n ths case s to mnmze, where s defned as z, z 0 (14) The rest of the constrants,.e. voltage lmts, feeders ampacty, load flow, etc., should also be added to the problem. 3. Smulaton Results In ths secton, wthout loss of generalty, the loss mnmzaton case s consdered. The objectve functon can be changed to voltage profle flattenng. 2.3 Test case In order to demonstrate the applcaton of the proposed algorthm, a test system s used here. The data for ths system s avalable n [20]. The system confguraton s reproduced here for reference n Fg. 4. Three DGs are present n ths system connected to Nodes 12, 20, and 54. Maxmum capacty of each unt s assumed to be 3.4 MVA, wth unty power factor (zero reactve power). Reactve power supports are avalable from two locatons by D-STATCOMs at Nodes 25 and 60. The lmts of reactve power exchange capabltes of these two unts are assumed to be ±2 MVAR. The voltage regulator connected between Nodes 6 and 7 has 20 tap postons, allowng the transformaton rato to vary between 0.9 and 1.1. A daly load curve s assumed for the feeder and all the nodal loads are scaled accordng to ths load curve shown n Fg. 5. DGs are assumed to be photovoltac type, wth the generaton pattern shown n Fg. 6. All loads are assumed to have the same voltage dependence characterstcs wth C 0.6, C 2.3. z z 7

8 Two cases are studed here and ther results are compared. The frst case assumes no reactve power support from the D-STATCOMs and no voltage regulator. The second case assumes both D-STATCOMs and the voltage regulator. The mnmum and maxmum nodal voltages are shown n Fg. 7. The maxmum voltages are above the lmt (1.04 p.u.) n almost half of the tme durng the day n the case wthout any control. By ntroducng the VVC algorthm, the voltages are bounded wthn the standard range. In case of losses, as shown n Fg. 8, about 27% reducton was acheved between 1pm and 3pm when the VVC routne s appled. The losses are slghtly hgher n hours that the generaton s not sgnfcant when VVC s appled. It s the result of brnng the voltages down wthn the standard lmt. Hgher voltages can lead to lower losses n the network n some cases. The reactve power drawn from the substaton, shown n Fg. 9, has substantally reduced due to the VAR support provded by the D-STATCOMs. The total actve power njecton by the substaton, gven n Fg. 10, dd not change sgnfcantly as was expected snce there were no changes n the loads or generaton. The only changes are the results of changes n losses and voltage-dependent loads. Fgure 11 shows the optmum tap postons when VVC s appled. Fgure 12 shows the reactve power exchange of the D-STATCOMs. As can be seen, durng the peak generaton tme, reactve power s absorbed by the second D-STATCOM to help lowerng the voltages. Also, the voltage regulator was set at low tap postons to reduce the voltage rse effect. Fgure 4. The 69-node test system wth DGs and DSTATCOMs locatons. Fgure 5. Daly load curve for the test system. 8

9 Fgure 6. Daly generaton curve for DGs (photovoltac system) Vmax_VVC Vmn_VVC Vmn_NC Vmax_NC Voltage (p.u.) Nodes Fgure 7. Mnmum and maxmum nodal voltages for No Control and VVC cases. Ploss (kw) Ploss_VVC Ploss_NC Tme (h) Fgure 8. Total losses for No Control and VVC cases. 9

10 Q Substaton (kvar) 3500 QS_VVC QS_NC Tme (h) Fgure 9. Total reactve power njecton by the substaton for No Control and VVC cases. P Substaton (kw) PS_VVC PS_NC Tme (h) Fgure 10. Total actve power njecton by the substaton for No Control and VVC cases. Fgure 11. Tap rato for the voltage regulator obtaned usng VVC algorthm. 7. Concluson A voltage-var control scheme s proposed n ths paper that allows for ncreasng the penetraton level of dstrbuted generaton wthout volatng the system operatonal lmts. In ths study, D- STATCOM was consdered as a source of reactve power support. However, t does not lmt the applcaton of the proposed algorthm to any other sources of VAR support, such as DGs, 10

11 2 1.5 DSTATCOM1 DSTATCOM2 Q Injecton (MVAR) Tme (h) Fgure 12. DSTATCOMs output obtaned usng VVC algorthm. batteres, swtchable capactor banks, etc. Ths algorthm also reduces the losses n the network and relves the reactve power drawn from the upper level sub-transmsson and transmsson level. A cost-beneft analyss would be requred to demonstrate the fnancal justfcaton of nstallng D-STATCOM n a dstrbuton feeder. Ths ncludes the captal and mantenance costs versus the revenues ganed by reducng losses and ncreasng the penetraton of clean energes. 8. References [1] Y. M. R. Dao, P. V. Etngov, S. Malhara, N. Zhou, R. T. Guttromson, J. Ma, P. Du, and N. S. C. Sastry, Analyss methodology for balancng authorty cooperaton n hgh penetraton of varable generaton, US Dep. Energy, [2] G. Strbac, A. Shakoor, M. Black, D. Pudjanto, and T. Bopp, Impact of wnd generaton on the operaton and development of the UK electrcty systems, Electr. Power Syst. Res., vol. 77, no. 9, pp , Jul [3] M. A. Ortega-Vazquez and D. S. Krschen, Estmatng the spnnng reserve requrements n systems wth sgnfcant wnd power generaton penetraton, Power Syst. IEEE Trans. On, vol. 24, no. 1, pp , [4] H. Ahmad and H. Ghasem, Securty-Constraned Unt Commtment wth Lnearzed System Frequency Lmt Constrants, IEEE Trans. Power Syst., vol. 29, no. 4, pp , July [5] P. P. Barker and R. W. de Mello, Determnng the mpact of dstrbuted generaton on power systems. I. Radal dstrbuton systems, n Power Engneerng Socety Summer Meetng, IEEE, 2000, vol. 3, pp [6] Canadan Standard Assocaton (CSA), Preferred voltage levels for AC systems, 0 to 50,000 V. Standard CAN3 C235-83, [7] P. M. Carvalho, P. F. Correa, and L. A. F. Ferrera, Dstrbuted reactve power generaton control for voltage rse mtgaton n dstrbuton networks, Power Syst. IEEE Trans. On, vol. 23, no. 2, pp , [8] H. M. Ayres, W. Fretas, M. C. De Almeda, and L. C. P. Da Slva, Method for determnng the maxmum allowable penetraton level of dstrbuted generaton wthout steady-state voltage volatons, IET Gener. Transm. Dstrb., vol. 4, no. 4, pp , [9] R. Tonkosk, D. Turcotte, and T. H. El-Fouly, Impact of hgh PV penetraton on voltage profles n resdental neghborhoods, Sustan. Energy IEEE Trans. On, vol. 3, no. 3, pp , [10] C.-H. Ln, W.-L. Hseh, C.-S. Chen, C.-T. Hsu, and T.-T. Ku, Optmzaton of photovoltac penetraton n dstrbuton systems consderng annual duraton curve of solar rradaton, Power Syst. IEEE Trans. On, vol. 27, no. 2, pp ,

12 [11] R. Tonkosk, L. A. Lopes, and T. H. El-Fouly, Coordnated actve power curtalment of grd connected PV nverters for overvoltage preventon, Sustan. Energy IEEE Trans. On, vol. 2, no. 2, pp , [12] H. Sughara, K. Yokoyama, O. Saek, K. Tsuj, and T. Funak, Economc and effcent voltage management usng customer-owned energy storage systems n a dstrbuton network wth hgh penetraton of photovoltac systems, Power Syst. IEEE Trans. On, vol. 28, no. 1, pp , [13] R. G. Wandhare and V. Agarwal, Reactve Power Capacty Enhancement of a PV-Grd System to Increase PV Penetraton Level n Smart Grd Scenaro. [14] X. Lu, A. Achhorn, L. Lu, and H. L, Coordnated control of dstrbuted energy storage system wth tap changer transformers for voltage rse mtgaton under hgh photovoltac penetraton, Smart Grd IEEE Trans. On, vol. 3, no. 2, pp , [15] A. R. D Fazo, G. Fusco, and M. Russo, Decentralzed control of dstrbuted generaton for voltage profle optmzaton n smart feeders, Smart Grd IEEE Trans. On, vol. 4, no. 3, pp , [16] C.-S. Chen, C.-H. Ln, W.-L. Hseh, C.-T. Hsu, and T.-T. Ku, Enhancement of PV penetraton wth DSTATCOM n tapower dstrbuton system, Power Syst. IEEE Trans. On, vol. 28, no. 2, pp , [17] J. R. Martí, H. Ahmad, and L. Bashualdo, Lnear power-flow formulaton based on a voltagedependent load model, Power Delv. IEEE Trans. On, vol. 28, no. 3, pp , [18] J. C. Das, Power system analyss: short-crcut load flow and harmoncs. CRC press, [19] X.-P. Zhang, C. Rehtanz, and B. Pal, Flexble AC Transmsson Systems: Modellng and Control: Modellng and Control. Sprnger Scence & Busness Meda, [20] J. S. Saver and D. Das, Impact of network reconfguraton on loss allocaton of radal dstrbuton systems, Power Delv. IEEE Trans. On, vol. 22, no. 4, pp , Acknowledgements Ths work was supported by Natural Scences and Engneerng Research Councl of Canada (NSERC) and MITACS-Accelerate fund No. IT Bography Hamed Ahmad receved the B.Sc. and M.Sc. degrees n electrcal engneerng from the Unversty of Tehran n 2009 and 2011, respectvely, and s currently a Ph.D. canddate n electrcal power engneerng at the Unversty of Brtsh Columba, Vancouver, BC, Canada. Hs research nterests nclude dstrbuton systems analyss, optmzaton algorthms, power system stablty and control, smart grds and hgh voltage engneerng. José R. Martí receved the Electrcal Engneerng degree from Central Unversty of Venezuela, Caracas, n 1971, the Master of Engneerng degree n electrc power (M.E.E.P.E.) from Rensselaer Polytechnc Insttute, Troy, NY, n 1974, and the Ph.D. degree n electrcal engneerng from the Unversty of Brtsh Columba, Vancouver, BC, Canada n He s known for hs contrbutons to the modelng of fast transents n large power networks, ncludng component models and soluton technques. Partcular emphass n recent years has been the development of dstrbuted computatonal solutons for real-tme smulaton of large systems and ntegrated multsystem solutons. He s a Professor of electrcal and computer engneerng at the Unversty of Brtsh Columba and a Regstered Professonal Engneer n the Provnce of Brtsh Columba, Canada. 12